Constitutional rules and party goals in coalition formation : an analysis of winning minority governments in Sweden

(1)

Constitutional Rules

and Party Goals in

Coalition Formation

An Analysis of Winning Minority

Governments in Sweden

Torbjörn Bergman

(2)

(3)

Constitutional Rules and Party Goals

in Coalition Formation

An Analysis of Winning Minority Governments in Sweden

AKADEMISK AVHANDLING

Som med vederbörligt tillstånd av rektorsämbetet vid Umeå universitet för vinnande av filosofie doktorsexamen framlägges till offentlig

granskning vid statsvetenskapliga institutionen Hörsal B, Samhällsvetarhuset

Fredagen den 31 mars 1995, kl 13.15 av

Torbjörn Bergman Fil kand

(4)

Constitutional Rules and Party Goals in Coalition Formation. An Analysis of

Winning Minority Governments in Sweden. Torbjörn Bergman, Department of Political Science, Umeå University, S-901 87 Umeå, Sweden.

ISSN 0349-0831, Research Report 1995:1

Distribution: Department of Political Science, Umeå University, S-901 87 Umeå, Sweden

Abstract

This study starts with two theoretical puzzles within the rational choice oriented literature on government formation in parliamentary democracies: the relative importance of constitutional rules and the existence of multiple party goals. From these puzzles stem the research questions that guide the study: First, what is the theoretical and empirical link between constitutional arrangements (including rules) and party goals? Second, what are the goals of political parties and how can these be studied? Third, relative to the goals of political parties and other constitutional arrangements, what is the importance of government formation rules for the empirical record of minority and majority governments?

Coalition theory provides the theoretical starting point from which the research questions stem. The historical-institutional strand of new institutionalism is used to guide the general understanding of the importance of institutional context. The rational choice oriented strand is used for a detailed study of the design of the Swedish government formation rules and an analysis of how the formation rules affect the goal s eeking (micro-logic) of actors.

Based on both cross-national data and an in-depth study of Swedish coalition and government formation, the analysis shows that the answer to research question number one is that the link between constitutional arrangemen ts and party goals is one of co-determination. The answer to research question number two is that party leaders pursue four main goals and that this should be an explicit model assumption. The answer to research question number three is that the government formation rules help determine the parties' bargaining positions and for that reason they are of significant importance for the formation of minority and majority governments.

Key words: Coalition theory, government formation, winning minority

governments, support party, parliamentary democracy, constitutional rules, constitutional design, new institutionalism, multiple party goals, expected utility

(5)

Constitutional Rules and Party G oals

in Coalition Formation

(6)

Constitutional Rules and Party Goals in Coali tion Formation. An Analysis of

Winning Minority Governments in Sweden. Torbjörn Bergman, Department of Political Science, Umeå University, S-901 87 Umeå, Sweden.

ISSN 0349-0831, Research Report 1995:1

Distribution: Department of Political Science, Umeå University, S-901 87 Umeå, Sweden

Abstract

This study starts with two theoretical puzzles within the rational choice oriented literature on government formation in parliamentary democracies: the relative importance of constitutional rules and the existence of multiple party goals. From these puzzles stem the research questions that guide the study: First, what is the theoretical and empirical link between constitutional arrangements (including rules) and party goals? Second, what are the goals of political parties and how can these be studied? Third, relative to the goals of political parties and other constitutional arrangements, what is the importance of government formation rules for the empirical record of minority and majority governments?

Coalition theory provides the theoretical starting point from which the research questions stem. The historical-institutional strand of new institutional ism is used to guide the general understanding of the importance of institutional context. The rational choice oriented strand is used for a detailed study of the design of the Swedish government formation rules and an analysis of how the formation rules affect the goal seeking (micro-logic) of actors.

Based on both cross-national data and an in-depth study of Swedish coalition and government formation, the analysis shows that the answer to research question number one is that the link between constitutional arrangements and party goals is one of co-determination. The answer to research question number two is that party leaders pursue four main goals and that this should be an explicit model assumption. The answer to research question number three is that the government formation rules help determine the parties' bargaining positions and for that reason they are of significant importance for the formation of minority and majority governments.

Key words: Coalition theory, government formation, winning minority

governments, support party, parliamentary democracy, constitutional rules, constitutional design, new institutionalism, multiple party goals, expected utility model, bargaining position.

(7)

Constitutional Rules and Party Goals

in Coalition Formation

An Analysis of Winning Minority Governments in Sweden

(8)

Department of Political Science Umeå University, Sweden Research Report 1995:1 ISSN 0349-0831 ISBN 91-7191-016-6 © Torbjörn Bergman

(9)

Acknowledgements

Much of the enjoyment I got from working on this study came from the intellectual puzzles that guided my work. First I was struck by the peculiar voting rule assumed in coalition theory. Later I realized the usefulness of trying to model multiple party goals. Since then my work has been about trying to study rules and party goals from the perspective of "new institutionalism".

By now, I owe a lot of thanks to a lot of people. Let me mention at least a few. At the Department of Political Science in Umeå, where the dissertation was finally completed, a small advising group composed of Svante Ersson, Bertil Hanson and Kjell Lundmark suggested significant improvements in the manuscript. I also got numerous helpful suggestions from Johan Eriksson, Rolf Hugoson, Anders Lidström, Sten Markgren, Håkan Myrlund, Lars Ricknell, Per Viklund and other participants of the department's weekly seminars. A special thanks to Professor Gunnel Gustafsson who during the completion of the manuscript proved the value of an advisor who is both critical and constructive at the same time. Stina Lindström guided me through technical matters concerning the final manuscript with both skill and friendliness.

Among the staff at the Department of Political Science in Minnesota, where this study first begun, I especially want to thank Professor Phil Shively who has an ability to point out one or two major flaws in a manuscript in a way that makes one eager to work on it again. Among Swedish political scientists, I want to mention Henrik Hermerén and Magnus Isberg for always being available when I realized that advice was needed. Not least, I want to thank the experts and politicians who allowed me interviews, often more than once. They helped me get behind the scene in my attempt to find out what really happened when governments were formed and the rules of the game designed. However, when I wrote this study I did not follow all of the advice I received and I am solely

(14)

grants from the British Economic and Social Science Research Council (E/00/223/002011), the Volkswagen Stiftung (11/38850) and the Nuffield Foundation (SOC205/125). Let me also gratefully acknowledge the support of the Department of Political Science at Umeå University; the Department of Political Science, University of Minnesota; the Fulbright Commission; Studentkårens Kempefond; Stiftelsen Siamon; Gösta Skoglunds Minnesfond and a fellowship at the Riksdag.

Parts of Chapter 3, Chapter 6 and Chapter 7 have been previously published as journal articles: "Multiple Goals and Constitutional Design: How the Swedish King Lost His Formal Powers", appeared in

Statsvetenskaplig Tidskrift, 1992, 95, 209-33. "Formation Rules and

Minority Governments", appeared in European Journal of Political

Research, 1993, 23, 55-66. "Constitutional Design and Government

Formation: The Expected Consequences of Negative Parliamentarism", was published in Scandinavian Political Studies, 1993, 16, 285-304.1 gratefully acknowledge that these parts are reprinted here with the permission of the publishers.

Chapter 4 is to some extent based on earlier work I conducted with Kaare Strom ("Sweden: Social Democratic Dominance in One Dimension", in Michael J. Laver and Ian Budge, eds. Party Policy and Government

Coalitions, 1992. London: MacMillan). I want to stress that Professor Strom

is in no way responsible for Chapter 4. I have independently updated, altered and favorably re-evaluated our earlier work.

Let me end by mentioning four people in particular. One of these is Professor Kaare Strom who was my mentor for years and is now a friend. Anders Östhol has been a constant source of inspiration for more than a decade. The third person is my brilliant wife, CK. Towards the end of the process of writing the dissertation, Jonathan made laughing easy. Thank you.

Umeå, Februari 1995

(15)

1

Constitutional Rules and Party Goals

Introduction

The formation of a new government is one of the most prominent recurring events of parliamentary democracy.1 Within political science it is common

to study government formation from the perspective of rational choice based coalition theory.2 This is a sub-field in which it is reasonable to speak

about cumulative progress.3 Early studies had little concern for rules and

institutions and conceptualized political parties as pursuing mainly one goal (office seeking). Today the leading literature incorporates both constitutional arrangements and multiple party goals (Budge and Laver 1986; Budge and Keman 1990; Laver and Schofield 1990, Laver and Budge 1992; Strom 1990b; Strom et al. 1994). Nonetheless, it is still the case that coalition theorists "have paid too little attention to the constitutional link between legislature and executive in European parliamentary systems" (Budge and Laver 1986, 488). In particular, this remains true with respect to existing rules for when a coalition is winning, i.e. when a new government can form.4

In formal coalition theory, following Riker (1962), a winning coalition has traditionally been defined as one which "contains over half of the membership or votes or weight in the decision-making system" (Riker 1962, 256). When they apply coalition theory to government formation in parliamentary democracies, Axelrod (1970, 170-83), Browne (1971, 394), De Swaan (1973, 130), Dodd (1974, 1105), Franklin and Mackie (1984, 673) and Robertson (1986, 537, 542-3) as well as Taylor and Laver (1973, 206-7) use Riker's (1962) definition of the size of a winning coalition.5

(16)

rule does not emerge from formal coalition theory, but was imposed on it by Riker (1962) as an universal institutional constraint.

The assumed voting rule has two parts. First, it holds that a winning coalition must "contain" members (in this case parties) who together control votes at or above a certain threshold. Second, it sets the minimum threshold at more than half of all members of the decision-making body. This is an absolute majority requirement.7 The institutional assumption (a winning

coalition must contain an absolute majority) is so restrictive that it excludes

minority governments.8 Thus, winning minority governments have been

anomalies in studies of government formation based on coalition theory (see the review articles by Browne and Franklin 1986; Laver 1986; Luebbert 1983). However, in cross-national comparisons of parliamentary democracies, about one-third of all governments turn out to be minority governments (Strom 1990b, 8). It is because of the assumed voting rule that winning minority governments have been anomalies.

Leading coalition theorists have learned to appreciate the flaws of the "must contain an absolute majority" criterion (Budge and Laver 1986; Budge and Keman 1990; Laver and Schofield 1990, Laver and Budge 1992; Strom 1990b; Strom et al. 1994). However, the continued lack of interest

in existing government formation rules is quite remarkable.9 In my

analysis, government formation rules are central. Yet unless one wants to argue that rules simply determine the outcome of government formation, which I do not, it is necessary to study how rules impact on the actors choices. To do so, an explicit model of party goals (i.e. the micro logic of the actors) is useful. But which goal is it that political parties pursue? They have been depicted as being predominately vote seekers (Downs 1957), office seekers (Riker 1962), and policy seekers (De Swaan 1973).10 Any

of these goals, and perhaps even others, can quite plausibly be thought of as a major goal of political parties.

My Research Questions

From this discussion I draw the following research questions: First, what

is the theoretical and empirical link between constitutional arrangements (including rules) and party goals? Second, what are the goals of political parties and how can these be studied? Third, relative to the goals of political parties and other constitutional arrangements, what is the

(17)

importance of government formation rules for the empirical record of minority and majority governments? The purpose of this study is to answer

these questions. This is a huge undertaking for any one investigation. My intention is therefore to provide well-grounded but tentative answers. They are well-grounded in the sense that I draw my conclusions from a series of analyses which all point strongly in the same direction. They are tentative in the sense that I do not claim to have "proven" that my answers are undisputable, only that there are good reasons to believe that they will hold true in further comparative investigation of government formation in parliamentary democracies.

I study party goals and the impact of constitutional arrangements within the general context of coalition and government formation. For reasons explained above, I find it particularly interesting to focus on the formation of minority governments. The topic of constitutional rules and party goals is relevant to other areas of study as well, but a discussion of this is left until the last chapter when I bring up the question of a future research agenda.

The outline of this chapter is as follows. In the next section I review the theoretical context from which I have drawn my research questions. This section also illustrates the logic of major theories of coalition formation. In section three I explain how my approach is different from the work of other scholars in the sub-field. In the fourth and final section I outline coming chapters.

Party Goals and Rules in Coalition Theory

Classical coalition theorists, such as Riker (1962), Axelrod (1970) and De Swaan (1973), explain coalition formation in terms of a rational choice of goal seeking actors. Primarily, political actors are assumed to be driven by one goal. In contrast to Downs's (1957) suggestion that political parties seek to maximize electoral returns (i.e. they are vote seekers), Riker (1962) proposed that an office seeking party does not have an incentive to maximize its share of the votes once it has reached a winning size. An office seeking party should not want to increase a coalition more than absolutely necessary. A larger coalition would only mean that each member

(18)

A minimum winning coalition is one which is no longer winning if it loses

one of its members (Riker 1962, 40).11

Riker (1962, 256) himself realized that his criterion that a winning coalition must contain an absolute majority was an auxiliary assumption.12

This criterion is really only appropriate if parties are pure office seekers.13

However, some coalition theorists have continued to use the "must contain an absolute majority" criterion even after most scholars relaxed the office seeking assumption and began to allow for a policy seeking goal (see, for example, Van Roozendaal 1993, 38; Van Deemen 1991, 143; Rasch 1993, 58). Schofield (1993, 17) states that, theoretically, other voting rules could be used, but he too uses the standard criterion of when a coalition is winning.

In a move probably driven by the recognition of the empirical frequency of minority governments, a second body of literature has begun to define winning in terms of the ability to get supporting votes from more than half of the members of parliament (rather than having to contain more than half of the members). For example, when discussing coalitions in parliamentary government formation, Austen-Smith and Banks (1990), Baron (1991; Baron 1993) and Ordeshook (1992, 303) discuss winning in these terms. That is, they allow for a minority coalition to be winning if it can get support from one or more parties outside of the coalition and thereby reach the absolute majority threshold. The assumed threshold still is an absolute majority, but because a (implicit) distinction is made between containing an absolute majority and having the support of an absolute majority, minority governments are no longer such a remarkable puzzle in coalition theory.

A third body of literature argues against the use of an absolute majority criterion in the first place. This literature has shown that formation rules in many countries do not require a winning coalition to have the votes of an absolute majority (Budge and Laver 1986; Laver 1986; Laver and Schofield 1990; Laver and Shepsle 1990a; Strom 1990b). This represents an important development in coalition theory.

In place of the absolute majority criterion, Budge and Laver have suggested a viability criterion,

A proto-coalition V will form a government if there is no alternative coalition A which is supported by parties controlling more legislative votes than does supporting V and which all supporters of A prefer to form rather than V (Budge and Laver 1986, 488).

(19)

What this means is that a coalition does not have to be actively supported by an absolute majority to form a government. Instead it can be sufficient if it can get a plurality of the votes in parliament, i.e. more votes than any other alternative. The viability criterion also implies that the spatial position of parties (i.e. policy seeking) should be taken into account when coalition formation is studied.

Axelrod (1970) and De Swaan (1973) were early proponents of the argument that not only winning office, but also policy seeking matters for coalition formation. De Swaan's argument is that an actor wants to join the winning coalition that is as close as possible to its own preferred policy position. Axelrod predicts that connected coalitions will form. Such a coalition is defined as one that "consists of adjacent members" in policy space (Axelrod 1970, 170). The introduction of a policy constraint challenges an assumption in early coalition theory. This is the assumption that all combinations of parties are feasible coalitions (Laver 1986, 34).

A Hypothetical Example

At this point an example can be helpful. Assume a five party system in which all five parties have an equal share of the seats in the parliament and name these parties from A to E. This is certainly an unrealistic example, but it is nonetheless a useful exercise. In a five party system, thirty-two (25)

combinations of parties are arithmetically possible (Axelrod, 1970, 166; Riker, 1962, 36). This set of thirty-two possible combinations includes one without any member.14 Excluding the empty coalition, but including those

with only one member as well as the coalition made up of all five parties, any one of thirty-one different governments can hypothetically form. Table 1.1 presents all possible combinations of parties and the coalitions predicted by the different theories.

Remember that a minimum winning coalition is one which is no longer winning if it loses one of its members (Riker 1962, 40). At any one time, a number of coalitions can meet this criteria.15 Nonetheless, Riker's

assumption of what constitutes a winning majority effectively limits the number of coalitions that are predicted. In this hypothetical example about one-third of all possible government combinations are predicted, i.e. they are in the prediction set. Almost half (fifteen) of the thirty-one possible

(20)

the prediction set on the grounds that they include one or more parties not necessary for the coalition to be of minimum winning size (i.e. they are oversized).16

Table 1.1 Prediction Set with Different Assumptions, a Hypothetical Example.

Government Size Prediction

Composition % Riker Axelrod Viabilitv Median L.

1 A 20 * 2 B 20 * 3 C 20 * * 4 D 20 * 5 E 20 * 6 A,B 40 * 7 A,C 40 8 A,D 40 9 A,E 40 10 B,C 40 * * 11 B,D 40 12 B,E 40 13 C,D 40 * * 14 C,E 40 15 D,E 40 * 16 A,B,C 60 * * * * 17 A,B,D 60 * 18 A,B,E 60 * 19 B,C,D 60 * * * * 20 B,C,E 60 * 21 C,D,E 60 * * * * 22 C,D,A 60 * 23 D,E,A 60 * 24 D,E,B 60 * 25 E,A,C 60 * 26 A,B,C,D 80 27 A,B,C,E 80 28 B,C,D,E 80 29 C,D,E,A 80 30 d,e,a,b 80 31 A,B,C,D,E 100 Predicted coalitions: 10 3 12 6

Note: The assumptions are explained in the text. The column based on the concept of the median legislator party is a subset of the column based on the viability criterion.

(21)

criterion that only coalitions between parties adjacent in policy space should form. Here it is assumed that the parties are aligned on one dimension from left to right in the order of A to E. The prediction set includes only three winning coalitions (A+B+C, B+C+D, C+D+E). With the traditional assumption that a winning coalition must contain an absolute majority, no minority governments are expected.

Let us now look at the viability criterion. Note that when the criterion that a winning coalition must contain an absolute majority is dropped, the prediction set becomes very large (i.e. it includes a large number of coalitions). Even with a rule that specifies that a coalition must have the support of an absolute majority, it is possible that a coalition which by itself holds only a relative small fraction of the seats in parliament can form a government.17 If parties B and C manage to agree to form a coalition

based on 40 percent of the seats in the parliament, and another party (party A) votes in favor of the coalition, perhaps in order to block a coalition of parties D and E, the coalition of B and C is winning. It is also possible that parties A and B vote for party C in order to block parties D and E. Thus, when the traditional criterion of winning size is dropped, even very small minority governments are hypothetically possible.

In its original formulation, the viability criterion assumed that a new government must have at least plurality support in the parliament. With colleagues, Budge and Laver (Laver and Schofield 1990, 66; Budge and Keman 1990, 34) have more recently reformulated the viability criterion from being able to win the votes of a plurality to one of being able to survive a vote of confidence (or no confidence) in the parliament. As I shall show in the next chapter, in many countries this is a more accurate institutional assumption. What the (new) viability criterion does, in effect, is to drop the "must contain an absolute majority" assumption while keeping the assumptions that (1) unless there are exogenous constraints on the process of coalition formation (such as a situation of national emergency), we should not expect oversized coalitions, and that (2) we should expect parties who are adjacent in policy space to form coalitions. As the third column from the left in Table 1.1 illustrates, with the viability criterion the new prediction set consists of twelve coalitions. Recall that under the traditional definition of a winning size and with a policy constraint (Axelrod), the prediction set included only three coalitions. Thus, with a

(22)

The fourth column illustrates a way to narrow the viability prediction set. The idea is that if parties form coalitions only among parties who are adjacent in policy space, party C will be decisive for the outcome of the government formation process. This is because party C is in control of the median legislator. The concept of a median legislator combines the size of the parties with their position on one dimension.18 The number of seats

that each party controls are added from Left to Right (or in the reverse order) until the party in control of the median legislator is found.

The fourth column in Table 1.1 illustrates the prediction set when it is expected that the party in control of the median legislator will be included in the government.19 By opposing, abstaining or by voting in favor of

some government alternative, party C will decide which combination of parties that forms the government. Because of this, only coalitions that include party C are likely to form. With these modification of early coalition theory, a prediction set which is likely to be both successful and efficient (i.e. relatively few of all possible combinations are predicted) is generated. In the example given in Table 1.1, the combination of the viability criterion and the median legislator expectation predicts that there are a group of six coalitions out of which one is likely to form a new government. This out of 31 hypothetical combinations.

The hypothetical example presented above schematically illustrates the logic of major theories in the coalition formation literature. It also shows that when the requirement that a winning coalition must contain an absolute majority is dropped, and replaced with a concern for viability and the median legislator, coalition theory can be both realistic and efficient. Empirically, it is also true that almost all governments include the median legislator party (Budge and Laver 1993).20

The empirical fact that almost all governments include the median legislator party leaves open the question of why it is that political parties sometime form minority governments and not majority governments. This is where the work of Strom (1984; 1985; 1986; 1990b) comes in. In what is probably the most authoritative work on minority governments, Kaare Strom (1990b) explain minority governments from the perspective of the multiple goals of political parties and the constitutional context of government formation. Political parties are assumed to have three goals, they are simultaneous office, policy and vote seekers. If a party has reason to believe that its voters are going to disapprove of a particular coalition, and also believes it can influence policy from its position in the parliament,

(23)

this decreases the party's desire to get into government. This can lead to coalition "avoidance". In such a case, a party might rather avoid joining a government coalition than entering one that it expects its voters to disapprove of.

Strom (1990b, 237-44) concludes that we should expect coalition avoidance in countries where elections are competitive and decisive for government formation and where the parliamentary opposition can be influential. In particular, it is the anticipation of future elections that is important when party leaders opt for minority governments and support parties avoid government coalitions. This challenges a prevailing notion in formal coalition theory that each coalition formation is an isolated event in which past history and expectations about the future has no place. Much of the key to the explanation of minority governments lies with the parties' expectations of future electoral fortunes and the extent to which the parties outside of the government can influence policy in the parliament.

In sum, my work draws upon earlier work on coalition theory and minority governments. Rather than an attempt to disprove this literature, the study constitutes an attempt to make a contribution to the sub-field's further development. My approach is explained in the next section.

My Approach

In this section I explain how I intend to answer the research questions posed above. I begin the study from a theoretical and comparative perspective. Empirically, I will also discuss the formation of minority governments in a comparative perspective. However, the empirical study is predominately focused on one country, Sweden. There are good reasons to study the case of Sweden.

A careful case study can serve the purpose of "theory building" (Eckstein 1975). By studying one case in detail, it is possible to discover nuances and relationships that could be lost, for example, in a cross-national statistical analysis. Zimmerman (1988) and Grofman (1989) have convincingly shown that within-country variation can be distinctively different from between-country variation. Because of this, for heuristic reasons it is useful to combine comparative knowledge with a focus on individual cases. Or in

(24)

351).

A good reason to focus on Sweden is that it has a high frequency of minority governments. This highlights the issue of when it is that a political party or a coalition is "winning". As mentioned above, in traditional coalition theory, minority governments were not seen as "winning" coalitions because of the institutional assumption of what constituted a winning size.

The most important reason for studying Sweden is that it provides an unusual opportunity to study the relationship between constitutional rules and party goals in a context in which constitutional rules were endogenous to the process of political decision-making. A new constitution was adopted in Sweden in the mid-1970s. Sweden thus provides a rare example of an encompassing constitutional revision in a stable representative democ

racy.21 Constitutional revisions have also occurred in Denmark and

Norway, two other countries with a high frequency of minority government, but these have not been as comprehensive as the adoption of a new Swedish constitution.

As a concrete case of constitutional design, I focus on what in Sweden is known as the Torekov Compromise (Torekovskompromissen). It was in Torekov, a small town in southern Sweden, that representatives of the Swedish parties agreed in August 1971 on the present design of the constitutional rules for government formation. In the analysis of the compromise in Torekov I show that the actors expected the government formation rules to have an impact on their own behavior and that their behavior has been consistent with these expectations.

The reasons behind the constitutional design and the bargaining position that it creates for the parties are striking examples of how constitutional rules and the goals of political parties are intertwined in coalition and government formation. I will argue that by facilitating tolerance for minority governments from parties outside the government, the Swedish government formation rules contribute to the high frequency of minority governments.

New Institutionalism

As mentioned above, most researchers studying coalition formation work from a rational choice perspective. With my concern for constitutional arrangements, I place myself within the research tradition of new

(25)

institutionalism. In the first part of the study I work within an analytical strand of new institutionalism that can be labelled "historical" institutionalism. In the latter part of the study I shift analytical perspective and work from a "rational choice" oriented strand. Below I explain why I have chosen to do so.

On a general level, the school of thought known as new institutionalism argues that we should search for explanations of political behavior in rules and institutions. In its most generic form the school represents a reaction against the 1950s and 1960s "behavioral revolution". At that time, behavioralists argued that scholars could not understand political behavior by studying formal rules such as those found in constitutions. They argued that there often is a large discrepancy between the power relations and functions described in constitutions and other organizational schemes and those that really exist and matter in political life. Their alternative was to study the behavior of groups and individuals rather than institutions. To the behavioral revolution the discipline owes rational choice, electoral studies, survey analysis and sophisticated quantitative analysis (for a discussion of these developments see, for example, Apter 1991; March and Olsen 1989; Olsen 1992; Putnam 1993).

When scholarly interest returned to institutions it was in an attempt to synthesize the behavioral tradition with the study of institutions.22

Institutionalist scholars such as March and Olsen (1984) do not deny that there can be a discrepancy between formal rules and political behavior. This, they argue, does not mean that organizational rules and the institutions of political life are unimportant. On the contrary, political behavior is shaped by both the formal and the informal institutions in which it exists. Putnam (1993, 8) points out that all strands of new institutionalism view institutions as "sticky", i.e. rules and other institutions are robust over time and costly to change. Once a particular set of institutional rules is chosen, this tends to determine future choices of institutional arrangements. Putnam speaks about this in terms of "path dependence", i.e. what comes before conditions what comes later. In this respect, earlier institutional arrangements provide a context which has an impact on present day choices. Beyond these similarities, however, there exist at least three main strands of new institutionalism. The three are sociological, (macro) historical and rational choice oriented.

(26)

social behavior is largely routine and originates from socialization, (3) by creating social roles institutions generate the actors understandings of themselves and (4) politics is as much about symbols and ways of understanding the world as it about making choices and decisions (see also Olsen 1985; Olsen 1992).23

While the sociological strand of new institutionalism is an interesting and challenging perspective, for the purpose of bringing constitutional rules into coalition theory it is less appropriate. I work on a different level of analysis. My questions about party goals and rules deal with individual (actor) choice in specified situations concerning specific decisions. A broader sociological approach to the study of government formation is beyond the scope of this study.24

The other two strands are more directly applicable. One of these is the historical (or Neo-Weberian) strand. Relative to the other strands, this represents the mainstream research tradition on institutions. It tends to emphasize the importance of the state and other institutional arrangements as something of an independent variable or as an actor (Skowronek 1982; Evans et al. 1985; Ikenberry et al. 1988; Steinmo et al. 1992). The analyses are often cast in terms of large-scale social processes such as social revolutions (Skocpol 1979) or the origins of neocorporatist arrangements (Rothstein 1992). For present purposes, with my focus on rules and other institutions, this strand guides attention towards the study of the impact of the broad constitutional arrangements on coalition and government formation.

The third strand of new institutionalism is the most applicable for bringing government formation rules into the analysis of coalition formation. The rational choice oriented strand is cast at a level of analysis that is appropriate for answering my empirical question. The "pure" rational choice tradition holds that one should understand social processes and outcomes in terms of the preferences and choices of actors, i.e. it is based on methodological individualism (see, for example, Buchanan and Tullock 1962, vi-vii; Van Roozendaal 1992, 3). The rational choice oriented strand of new institutionalism is placed firmly in the ontological and epistemological tradition of methodological individualism. The choices of actors are always the basis for explanation. Because of its focus on goal oriented actors, the third strand might perhaps be seen as a new version of rational choice theory.

(27)

institutionalism. One reason is that this leads to an explicit and immediate recognition of the importance of rules and other institutions. The difference between a rational choice model in which attention is given to rules and the third strand of new institutionalism is foremost a difference of emphasis, but an important one. In the former case, rules and institutions have the character of an afterthought, in the latter case they are a starting point. In contrast to pure rational choice theory, the rational choice strand of new institutionalism searches for the explanation of behavior in the relation between actors' goals and institutions. For scholars working in this tradition, rules constrain the range of possible behavior and they favor certain outcomes over other (Przeworski 1988; Riker 1980; Shepsle 1986; Shepsle 1989; Tsebelis 1990).

I also prefer to speak about new institutionalism because one of the most fascinating characteristics of all three strands is that authors in this tradition ask questions about the origins of institutions. And at least within the historical and the rational choice oriented strands, scholars tend to emphasize that rules and institutions should be seen as designed rather than self-evident or evolutionary (Knight 1992; Shepsle 1989; Mershon 1994). The impact of institutions might not always be those expected, but the starting point is that a better understanding of the importance of institutional design is necessary for an improved understanding of political life.

In the remainder of this chapter I present the outline of coming chapters.

The Outline of the Study

The use of two distinct analytical perspectives is reflected in the presentation of the study. The study consists of two main parts. In Chapters 3 and 4 I combine the historical strand of new institutionalism with coalition theory in a study of the empirical record of coalition and government formation. In this part of the study, party goals and existing constitutional rules are taken as given, i.e. their origins are not discussed. In the second part (Chapters 5 through 8), in an attempt to move beyond the argument presented in earlier chapters, I shift to the analytical perspective of the rational choice oriented strand of new institutionalism. I combine an explicit model of multiple goals and arenas with a perspective

(28)

can be used to study constitutional design, i.e. the process of the creation of new constitutional rules. In the final empirical chapter, I reverse the relationship under consideration and explain how the resulting constitutional rules help determine the parties' bargaining positions. One consequence of this research design is that I discuss theory not only in this chapter but also in Chapter 5 (and later). Together the two parts of the study provide the basis for the answers to my research questions. They also provide a basis for an outline of a future research agenda.25

In terms of individual chapters, the study is organized as follows. In Chapter 2 I discuss definitions of concepts and terms. I discuss methods and sources and I explain a number of choices that I have made when designing the study. It is also here that I discuss complementary approaches and the possible pitfalls of working from the approach that I have chosen.

From a comparative perspective, in Chapter 3 I show that minority governments, and in particular small minority governments, are more frequent in countries that share a particular kind of government formation rule. In these countries, the formation process is, on average, also much shorter. These findings are not accounted for by current coalition theory.

In Chapter 4 I present the empirical record of Swedish coalition and government formation in the post-war era. Chapter 4 also provides an initial explanation of this historical record. Constitutional rules, voter alignment and party positions determine much of coalition and government formation. This provides a necessary background for the more direct answers to my research questions.

In Chapter 5 I develop, explain and defend the multiple goals model. Theoretically, I connect each of the goals to a major arena for party behavior. I use this model, in Chapters 6 and 7, to study the party goals behind the constitutional design of the present Swedish government formation rules. In Chapter 6 I provide an example of how, when an issue is (potentially) electorally salient, both the reasoning and the behavior of party leaders is consistent with the existence of multiple goals in multiple arenas. The concrete issue deals with the almost complete elimination of the Swedish Head of State's (the King's) formal powers in the early 1970s. This particular example gives the context in which the present Swedish formation rules were chosen.

In Chapter 7 I explain why the present Swedish constitution requires the Riksdag to vote on a new government. I will also show that the actors involved in the design of the government formation rules expected the rules

(29)

to have consequences for the outcome of the government formation process. By not creating a minimum threshold for the election of a new government, the formation rules were supposed to facilitate a short formation process and the formation of minority governments. The outcome of the government formation process has been consistent with those expectations. Theoretically, another main point of this chapter is that it provides an illustration of the limits of the goal seeking model. The analysis of constitutional design suggests that a goal seeking explanation is of great use for the study of some kinds of decisions, but of limited use for the study of others.

In Chapter 8 I focus on the link between the formation rules and the empirical record of government formation. Here I explain why minority governments are frequent, why very small minority governments can form, and why the formation process is short in the context of a particular constitutional design of government formation rules. I use the model to show how the design of the government formation rules help determine the parties' bargaining positions.

Chapter 8 is more speculative than other chapters. This is because it involves a significant amount of counter-factual reasoning. I ask questions about what would have happened if the rules had been different. This is an unusual approach in political science. To establish a more solid link between formation rules and party goals in terms of the micro-logic of the actors, one must conduct comparative studies. This would make possible an empirical study of the impact of very different formation rules. However, in order to usefully conduct such comparative studies it is necessary to have a theoretical foundation from which to start. Chapter 8 should be seen in this light. It constitutes an attempt at theory building for the purpose of establishing a guide for future research on the topic.

In Chapter 9 I summarize the answers to my research questions and I draw out the theoretical and empirical implications of these answers. I also address the issue of how to study the link between constitutional rules and party goals in further research.

(30)

Notes

1. A parliamentary democracy is a representative democracy in which the government (the executive) must be supported, or at least tolerated, by the parliament (Bogdanor 1984; Brusewitz 1929). In keeping with much of the literature I have chosen to use the term government rather than cabinet to denote the executive in a parliamentary democracy. A presidential system is considered to be the other "pure" form of representative democracy. In a presidential system, the process of forming the executive is institutionally separate from the process of filling seats in the assembly, because both branches are popularly elected. There exist also "mixed" forms of legislature-executive relations (Shugart and Carey 1992, 2). France (Fifth Republic) is an often used example of a mixed system.

2. Of course, government formation is not only interesting from the perspective of coalition theory. Government formation matters both as an important aspect of representative democracy and because of its potential impact on political economy. Minority governments tend to have shorter tenure in office than majority governments. However, while most governments lose votes in subsequent elections, minority governments suffer smaller losses in elections than both majority parties and multiple-party majority coalitions. Multi-party majority coalitions is the government type that tends to suffer the greatest vote losses at election time (Strom 1990b, 113-31). In countries with majoritarian electoral systems, left-wing or right-wing party control of the government has an observable impact on budgetary spending patterns. With proportional elections and minority governments, the ideological position of government parties tend to have a marginal partisan effect on budgetary spending patterns (Biais et al. 1993; Hofferbert and Budge 1992, Jackman 1986; Petry 1991; Klingemann et al. in press). Thus, consistent with the argument in this study, in such systems parties need not necessarily be in government to exercise influence on national policy-making. In cross-national OECD data, there exists some statistical evidence that national budget deficits are negatively correlated with the duration of governments. When government duration increases, the probability of a long-term budget deficit decreases (Persson 1993; Biais et al. 1993, 55; Edin and Ohlsson 1990 on Swedish data). There exists, on the other hand, no evidence that minority governments must have negative effects on budget balances. In fact, in the case of Sweden, Molander (1992, 28) concludes that stable minority governments before the 1970s had a positive effect on the state's budget balance.

3. The rational choice approach to coalition formation constitutes a research program (Lakatos 1970) in political science. In turn, rational choice theory belongs to a major type of explanation, namely one which explains social phenomena on the grounds of purposeful action. Some other forms of explanations are causal (or correlational), functional and dialectic explanations (Elster 1986). In short, causal explanations explain variations in one social variable on the grounds of variation in other variables. Functional explanations explain social phenomena in terms of the role they perform in a larger context (system). Dialectic explanations explain social phenomena in terms of the conflicting forces (material or idea-based) that produce them.

(31)

4. A theoretical point often left unclear in the literature is that what it means to be "winning" depends on the goal(s) of the actor (on this point, see also Schlesingers 1991, 138). For example, if a party wants to get into national office, being successful in terms of electoral votes can be seen as a "non-win" if the party does not get into government. However, for the purpose of this study, the term winning will be used in only two ways. Either it refers to a party or a coalition of parties forming a government (the national executive) or it refers to a coalition which, in coalition theory, belongs to the prediction set. This narrow use of the concept of winning is motivated by the fact that this is the way it has been used within the coalition formation literature.

5. The size of a party or a coalition is measured by its share of the parliamentary seats.

6. Taylor and Laver (1973, 229) also test coalition theories with other absolute thresholds (49, 48, 45 and 40 percent of the seats). Because some minority governments are now considered to be winning, this improves the success of coalition theory. However, such absolute voting rules do not exist in any parliamentary democracy included in their data.

7. There exists some international disagreement about the meaning of the prefixes associated with the term majority. In Bogdanor's (1987, 350) political dictionary, absolute majority is defined as a "majority of all votes cast" and relative majority as having "more votes" than any other alternative. This definition does not recognize the distinction between a majority of all members and a majority of all votes cast. From the perspective of coalition theory, this is an important distinction. The former requires a winning coalition to have the support of a specific number of members of parliament. The latter only requires a winning coalition to have the majority among the members who actually do vote. Thus, in this study (as in Riker 1962), an absolute majority is a majority of all the members of parliament. A

relative majority is a majority of the members of parliament who actually vote on an issue (i.e. this excludes abstentions and absentees). What Bogdanor (1987) calls a relative majority I consider to be a plurality rule. That is, a proposal is winning when it gets more votes than any other alternative (without having to encompass a relative majority of the votes cast). Note that the assumed voting rules contrasts with the voting rule used for most decisions in most parliaments around the world. A comprehensive survey of parliaments concludes that: "ordinarily, decisions are taken by a majority of the votes cast" (Parliaments of the World 1986, 516). Thus, in parliamentary decision-making the most common voting rule is relative majority. 8. A minority government is a government which holds less than half of the seats in

the parliament (Strom 1990b, 6-7).

9. The literature does point to the potential importance of differences in government formation rules. One example is Strom (1990b, 79) who points out that the presence of a constitutional requirement of investiture (i.e. parliamentary approval)

(32)

governments (Strom 1990b, 110). However, as is the case with his colleagues, the full consequences of government formation rules are beyond the scope of Strom's analysis.

10. The political party is the "unit of analysis" which forms or does not form governments. Scholars such as Downs (1957) and Sartori (1976, 63) hold that the defining feature of a political party in a representative democracy is the fact that it participates in elections. Following this, my working definition of a political party is that it is an assembly of men and women who take part in an general election. However, as argued in the text, vote seeking is not the only goal of political parties. Perhaps this marks an important difference between the firm in economics and the party in political science, the former can more plausibly be said to seek mainly one goal (profit).

11. Riker's (1962) theory of minimum winning coalitions is probably the most widely known of all coalition theories. It reads,

In n-person, zero-sum games, where side-payments are permitted, where players are rational, and where they have perfect information, only minimum winning coalitions occur (Riker 1962, 32).

Riker rephrased this "size principle" into a "sociological law",

In social situations similar to n-person, zero-sum games with side-payments, participants create coalitions just as large as they believe will ensure winning and no larger (1962, 32-33).

Riker's (1962) theory of minimum winning coalitions corresponds to a prediction set—the V-set. In the vocabulary of the rational choice tradition, a minimum winning coalition is an equilibrium outcome of a cooperative game. Traditional coalition theory is of this kind. In a cooperative game, an equilibrium "corresponds to an outcome in which no coalition has the incentive or the means for unilaterally insuring an improvement in the welfare of all its members" (Ordeshook 1982, 26). 12. Riker was also one of the leading scholars in the rational choice sub-field of social

choice. Social choice theory is concerned with the consequences of different voting rules, some of which are plurality, plurality runoff, approval, Black, Borda, Coombs and Hare voting procedures (Hermansson 1990; Lane and Berg 1991; Nurmi 1987). However, to my knowledge Riker never brought the full implications of social choice theory into coalition theory. In his work on voting rules within social choice, Riker (1982, 41-64) pointed out that in a binary choice (i.e. in a choice between two alternatives), the relative majority voting rule is the only voting method which satisfies the three properties of monotonicity, undifferentiatedness and neutrality. Monotonicity means that if more voters become in favor of a particular alternative, the voting rule is such that this alternative has a better chance of winning. Undifferentiatedness means that all votes are equally important. And, finally, neutrality means that a voting rule does not favor any alternative (Riker 1982, 56-9). According to Riker (1982, 57; Riker 1986, 104) an absolute majority rule violates the neutrality property. With an absolute majority rule, a motion only wins if it receives a certain number of votes, in all other cases it loses. The voting rule favors one alternative (in this case the side opposed to the motion) and discriminates against the motion. Thus, the threshold at which a coalition

(33)

traditionally has been assumed to be winning (i.e. must contain an absolute majority) violates the third property of the relative majority rule, i.e. it is not a neutral voting rule. As I will show, the use of non-neutral voting rules impacts on coalition formation.

13. Or, to be more precise, Riker's criterion is only appropriate if (1) political parties are modelled as pure office-seekers and (2) the game is constant-sum (zero-sum). For a discussion of the latter, see Budge and Laver (1986, 487-90). If, as Riker (1962) assumed, it is true that political parties are only (or pure) office seekers, there is no reason to believe that they will tolerate a government in which they are not included (Budge and Keman 1990, 49). There would be no reason to expect minority governments. Unless a coalition contains members in control of an absolute majority, the opposition parties will simply join forces and put themselves in government (Budge and Keman 1990, 11).

14. In practice the "empty" coalition can be said to correspond to a non-partisan government. In the post-war period, such governments have formed in both Finland and Portugal (Strom 1990b, 58).

15. Gamson (1961) narrows the prediction set with the expectation that the coalition which among all winning coalitions is in control of the smallest possible number of seats will win. Leiserson (1970) has suggested that the coalition that can ensure an absolute majority (50%+1) with the fewest number of member parties should be the predicted one. In practice, these alternative definitions can limit the prediction set. However, because the parties in the example are all of equal size, in Gamson's and Leiserson's definitions generate prediction sets that are identical to Riker's. 16. According to Riker's (1962) theory, a coalition that contains more members

(parties) than are needed to ensure an absolute majority is an "oversized" coalition. However, Laver argues that oversized coalitions are not much of a theoretical problem,

This situation may arise, among other reasons, because constitutional amendments requiring a qualified majority are on the policy agenda, or because party discipline is poor and solid party votes in the legislature cannot be guaranteed, even on crucial issues. Here we are dealing with the familiar concept of the necessary 'working majority' for a government (Laver 1986, 38).

Budge and Keman (1990) stress that such oversized coalition are likely to occur when the very foundation of representative democracy is threatened. That is, coalition formation in times of significant external pressure is likely to be different from the process of coalition formation in more normal times.

17. As I will show in Chapter 3, the criterion that a new government must have the support of an absolute majority (without having to contain such a majority) is used in a few parliamentary democracies. Of the 15 democracies I study, this includes Israel, Germany and Spain.

(34)

18. A corresponding concept is that of the "core" of the bargaining game. If there is only one salient dimension, the median legislator is identical to the core. The core is a position in policy space which cannot be defeated and which can defeat all other. However, in multi-dimensional space the core is often empty, i.e. no coalition of parties has such a position. With three or more dimensions, the core is almost always empty. This is known as the chaos theorem (Laver and Schofield 1990, 121-9). Such results have been conducive to the rational choice version of new institutionalism which explains the discrepancy between the theoretical expectation of instability and the empirical observation of relative stability by emphasizing the stabilization induced by institutional arrangements (Laver and Budge 1992, 2-4). Schofield (1993) has suggested "the heart", a subset of policy space, as another a promising solution concept. However, the sophisticated formal modelling that underlies that concept still relies on the institutional assumption that only a coalition with more than half of the seats is a winning coalition. As Laver and Schofield (1990, 128) suggest, perhaps studying the effects of real-world institutional decision rules can shed light on the relationship between theoretically unstable policy positions and relatively stable real-world government formations. This study attempts to take a small step in that direction.

19. The combination of the viability and the median legislator criteria leads to a prediction set which assumes (1) that a winning government consists of one party or a group of parties that are connected in policy space; (2) includes the median legislator party; (3) has no minimum size requirement but (4) excludes coalitions that consists of a party not needed once an absolute majority threshold is reached. 20. There are some exceptions to this rule. For example, Str0m and Leipart (1993) have shown that five out of 17 Norwegian governments did not include the median legislator party. For four out of these five governments, the median legislator party belonged to an alternative coalition of parties which had just left a failed government. In the fifth case, the median legislator party did not join the government because of an electoral pre-commitment which was not met. The party, in this case the Center Party, had promised to join the Conservatives in government only as part of a larger coalition which failed to materialize. This study is not about the anomaly of a party (or a coalition) who is winning without being in control of the median legislator. While this is an intererting area for further research, for

present purposes a theoretical understanding of why median legislator parties usually have a privileged bargaining position is sufficient.

21. Constitutional design continues to be a common theme in the Swedish public debate. One example is a highly publicized report commissioned by the government in 1992. In their 1993 report the members of the commission, five economists and one political scientist, argued that many of the problems in the Swedish economy (budget deficit, a financial system in crisis, rising unemployment, etc.) are rooted in political instability. As one remedy they advocated reforms to increase government authority relative to the Riksdag. To facilitate this they suggested among other things that Sweden combine a German type "constructive vote of no confidence" with a French inspired legislative option that enables the government to pass legislation through the parliament unless a vote of no confidence is passed

(35)

against it (SOU 1993:16, 150-68).

22. The trends discussed in the text are about broad shifts and developments within the social sciences in general and political science in particular. It is in contrast to the relative lack of attention given institutions in the 1960s and 1970s that "new institutionalism" represents a shift of focus within the discipline. Of course, the

impact of institutions and the design of institutions has for centuries fascinated political philosophers and scholars. In the context of Swedish constitutional politics, excellent examples of scholars sharing similar concerns without using the label of new institutionalism are Back (1980), Board (1980) and Verney (1957).

23. Putnam ( 1993, 7) has labelled this the perspective of "organization theorists" within new institutionalism. For a brief introduction to the sociological strand of new institutionalism, see DiMaggio and Powell (1991, 8-11).

24. However, the sociological strand of new institutionalism is close to culture theory and as such it can be said to enter the analysis through the issue of preference formation (see Chapter 5).

25. After all, as Dunleavy has so aptly put it,

We still know so little about the mechanisms and operations of democratic control, government power and state institutions-in a world where the number of liberal democracies is growing rapidly, and the striving for democracy is demonstrably near-universal (Dunleavy 1991, 259).

(36)

2 Potential Problems and Methodological Choices

Introduction

I begin this chapter with a discussion of two fundamental issues about the approach that I use: the existence of goal seeking and the potential for institutional determinism. In the third section I discuss complementary approaches to the study of government formation. From there I move on to a more technical discussion of concepts and terms (Section four) as well as methods and sources (Section five).

Two Fundamental Issues

As explained, my analysis starts from a combination of coalition theory and the historical strand of new institutionalism. From this perspective parties are assumed to have goals and a central task is to look for the importance of rules and institutions. In the second part of the study, from the perspective of the rational choice oriented strand of new institutionalism, party goals and rules are even more in focus. For reasons explained above, I highlight government formation rules. For reasons also explained above, the formation of winning minority governments is of particular interest in the empirical study.

On a fundamental level, one can argue about whether or not political parties seek goals and whether a focus on rules will lead to institutional determinism. The purpose of this section is to explain why I believe that a goal seeking approach has merit and that I do not want to reduce the analysis to institutional determinism.

Constitutional rules and party goals in coalition formation : an analysis of winning minority governments in Sweden

Constitutional Rules

and Party Goals in

Coalition Formation

An Analysis of Winning Minority

Governments in Sweden

Torbjörn Bergman

Constitutional Rules and Party Goals

in Coalition Formation

Constitutional Rules and Party G oals

in Coalition Formation

Constitutional Rules and Party Goals

in Coalition Formation

An Analysis of Winning Minority Governments in Sweden

Contents

Acknowledgements

Constitutional Rules and Party Goals

2

Potential Problems and Methodological Choices